GALOIS POINTS ON VARIETIES MOSHE JARDEN AND BJORN POONEN Abstract. A field K is ample if for every geometrically integral K-variety V with a smooth K-point, V (K) is Zariski dense in V . A field K is Galois-potent if every geometrically integral K-variety has a closed point whose residue field is Galois over K. We prove that every ample field is Galois-potent. But we construct also non-ample Galois-potent fields; in fact, every field has a regular extension with these properties. 1. Introduction Definition 1.1. Let X be a variety over a field K. By a Galois point on X, we mean a closed point whose residue field is Galois over K. We say that K is Galois-potent if every geometrically integral K-variety has a Galois point. Question 1.2. Is every field Galois-potent? We do not even know if Q is Galois-potent. On the other hand, we have the following definition of Florian Pop: Definition 1.3 (cf. [Pop96, p. 2] and [Jar11, Chapter 5]). A field K is called ample (or large or anti-Mordellic) if for every geometrically integral K-variety V with a smooth K-point, V (K) is Zariski dense in V . Pseudo-algebraically closed (PAC) fields and Henselian fields (e.g., Qp ) are ample; see [Jar11, Chapter 5] for these and many more examples. Our first main theorem is the following: Theorem 1.4. Ample fields are Galois-potent. Some non-ample fields too are Galois-potent. For example, finite fields are non-ample, but have abelian absolute Galois group and hence trivially are Galois-potent. Less trivially, we can also construct infinite non-ample fields that are Galois-potent: in Section 5 we will prove the following. Theorem 1.5. Every field admits a regular extension that is Galois-potent but not ample. Bary-Soroker and Fehm in [BSF13, Section 2.2] write that “all infinite non-ample fields appearing in the literature are Hilbertian”. The non-ample fields we construct in the proof of Theorem 1.5 turn out to be Hilbertian too (Remark 5.4), and in particular they have non-abelian absolute Galois group. Date: November 26, 2015. 2010 Mathematics Subject Classification. Primary 12E30; Secondary 11G35, 14G99. The first author was supported by the Minkowski Center for Geometry at Tel Aviv University, established by the Minerva Foundation. The second author was supported in part by National Science Foundation grant DMS-1069236 and a grant from the Simons Foundation (#340694 to Bjorn Poonen). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the funding agencies. 1 Question 1.6. Are there also infinite non-ample fields with abelian absolute Galois group? If the answer to Question 1.6 were yes, it would immediately yield another example of a non-ample Galois-potent field. But there are several questions and conjectures in the literature that each imply a negative answer to Question 1.6, as we now explain. Let GK be the absolute Galois group of K. • Is every infinite non-ample field Hilbertian? (Cf. [BSF13, Section 2.2] again.) A Hilbertian field K cannot have an abelian GK . • Is every infinite field K with topologically finitely generated GK ample? (See the paragraph before Theorem 5.5 of [JK10], or see [BSF13, 4.2, Question III].) By [Koe01], if GK is abelian, either K is henselian and hence ample, or GK is procyclic and in particular topologically finitely generated, and hence ample if the question at the start of this bulleted item has a positive answer. • Is every infinite field with prosolvable GK ample? (See [BSF13, 4.1, Question II].) If so, then every infinite field with abelian GK is ample too. 2. Notation Let K be a field. Let K be an algebraic closure of K. By a K-variety we mean a separated scheme of finite type over K. Given a K-variety X and a finite extension L ⊃ K, let XL be the L-variety X ×K L; similarly, if φ is a K-morphism, let φL be its base change. If X is an integral K-variety, let K(X) denote its function field. By a K-curve, we mean a smooth geometrically integral K-variety of dimension 1. By the absolute genus gX of a K-curve X, we mean the genus of the smooth projective model of XK . Given a curve X and n ≥ 1, let X (n) denote the nth symmetric power, defined as the (variety) quotient of X n by the symmetric group Sn . 3. Restriction of scalars Let K ⊆ L be a finite extension of fields. For any quasi-projective L-variety X, the functor sending each K-scheme S to the set X(SL ) is representable by a K-variety ResL/K X called the restriction of scalars (cf. [BLR90, §7.6, Theorem 4]). If, in addition, L is separable over K, then ResL/K X after base field extension becomes isomorphic to a product of [L : K] conjugates of X; in particular, if X is geometrically integral and smooth of dimension d over L, then ResL/K X is geometrically integral and smooth of dimension [L : K]d over K. 4. Ample fields are Galois-potent In this section we prove Theorem 1.4. Proposition 4.1. Let K be an ample field. Let X be a geometrically integral K-variety. Let N be a finite separable extension of K. If X has a smooth N -point, then X has a point with residue field N . Proof. Replacing X by a Zariski-open subvariety, we may assume that X is smooth and affine. For each finite separable extension L of K, define X L := ResL/K (XL ). Thus X L (S) = XL (SL ) = X(SL ) for any K-scheme S, and in particular, X L (K) = X(L). If L ⊆ L0 are two 0 such extensions, the natural map X(SL ) → X(SL0 ) can be rewritten as X L (S) → X L (S), 0 and it is functorial in S, so Yoneda’s lemma yields a K-morphism X L → X L . 2 We have X N (K) = X(N ), which is nonempty by assumption. Also, K is ample, and X N is smooth and geometrically integral, so X N (K) is Zariski-dense in X N . There are only finitely many fields L with K ⊆ L ( N , and for each such L, we have dim X L < dim X N . Thus X N has a K-point outside the image of every morphism X L → X N . In other words, X(N ) has a element outside X(L) for every L. For this element, the image of Spec N → X is a closed point with residue field N . Proof of Theorem 1.4. Let K be an ample field, and let X be any geometrically integral K-variety. Choose a finite separable extension N of K such that X(N ) 6= ∅. Enlarge N to assume that N is Galois over K. By Proposition 4.1, X has a closed point with residue field N. 5. Infinite Galois-potent fields that are not ample In this section we prove Theorem 1.5. Lemma 5.1. Let K be an algebraically closed field. Let X, Y, C be K-curves with gC > 1. Any rational map X × Y 99K C factors through one of the projections to X or Y . Proof. A rational map φ : X × Y 99K C may be viewed as an algebraic family of rational maps X 99K C parametrized by (an open subvariety of) Y . But the de Franchis–Severi theorem [Sam66, Théorème 1] implies that there are no nonconstant algebraic families of nonconstant rational maps X 99K C. Thus either the rational maps in the family are all the same, in which case φ factors through the first projection, or each rational map in the family is constant, in which case φ factors through the second projection. Lemma 5.2. Let X be a curve over a field K. Let F = K(X (2) ). Then X has a point over a quadratic extension of F , and C(F ) = C(K) for every K-curve C of absolute genus > 1. Proof. Either projection X × X → X gives a point of X over the quadratic extension K(X × X) of K(X (2) ) = F . Let c ∈ C(F ). Then c corresponds to a rational map X (2) 99K C. Composing X ×X → X (2) with such a rational map yields a rational map φ : X × X 99K C. By Lemma 5.1, φK is constant on all vertical copies of XK or constant on all horizontal copies of XK . Since φK is S2 -invariant, it is constant on all vertical and horizontal copies of XK . Thus φK is constant. Hence φ is constant. Equivalently, c ∈ C(K). Lemma 5.3. Every field K admits a regular extension K 0 such that (i) Every K-curve X has a point over an at most quadratic extension of K 0 (possibly depending on X). (ii) For every K-curve C of absolute genus > 1, we have C(K 0 ) = C(K). Proof. Let (Xα )α<τ be a well-ordering of the set of K-curves up to isomorphism, indexed by an ordinal τ . For α ≤ τ , define Kα by transfinite induction as follows: • Let K0 := K. (2) • For each α < τ , define Kα+1 := Kα (Xα ); • If α is a limit ordinal, define Kα := limβ<α Kβ . −→ Let K 0 := Kτ . Then K 0 is regular over K by [FJ08, Corollary 2.6.5(d)], whose proof remains valid when the cardinal m is replaced by an ordinal τ . 3 (i) Any K-curve X is Xα for some α < τ . Lemma 5.2 shows that X has a point over a quadratic extension of Kα+1 , and hence also a point over an at most quadratic extension of K 0 . (ii) Let C be a K-curve of absolute genus > 1. By Lemma 5.2 and induction, C(Kα ) = C(K) for all α ≤ τ . In particular, C(K 0 ) = C(K). Proof of Theorem 1.5. Let K be the given field. Construct a sequence of fields L0 , L1 , . . . inductively as follows: let L0 := K(t) with t transcendental over K, and let Li+1 be the regular extension of Li given by Lemma 5.3. Let L∞ := lim Li . −→ Let X be an L∞ -curve. Then X is definable over some Li . The conditions in Lemma 5.3 imply that X has a point over an at most quadratic extension of Li+1 , and hence a point over an at most quadratic extension of L∞ . Every geometrically integral L∞ -variety other than a point contains an L∞ -curve, so L∞ is Galois-potent. Choose a non-isotrivial curve C of absolute genus > 1 over L0 = K(t) such that C has a smooth L0 -point; then C(L0 ) is finite [Sam66, Théorème 4]. The conditions in Lemma 5.3 imply that C(L0 ) = C(L1 ) = · · ·, so C(L∞ ) = C(L0 ), which is finite. Thus L∞ is not ample. Remark 5.4. The field L∞ constructed in the proof of Theorem 1.5 is Hilbertian, because of the following two facts: (a) Any finitely generated transcendental extension of a field is Hilbertian [FJ08, Theorem 13.4.2]. (b) If (Kα )α≤τ is an ascending tower of fields indexed by an ordinal τ , and Kα is Hilbertian for each α < τ , and Kα+1 is a regular extension of Kα for each α < τ , and Kα = limβ<α Kβ −→ for each limit ordinal α ≤ τ , then Kτ is Hilbertian. (Fact (b) appears as [FJ08, Chapter 12, Exercise 2], but the hypothesis Kα = limβ<α Kβ is −→ missing there.) Proof of (b). Given irreducible polynomials fi (~t, x) over Kτ for i = 1, . . . , r, each separable in x, we must find infinitely many ~a over Kτ such that the specializations fi (~a, x) for i = 1, . . . , r are irreducible over Kτ . There exists α < τ such that all the fi are defined over Kα . Since Kα is Hilbertian, there exist infinitely many ~a over Kα such that fi (~a, x) for each i is irreducible over Kα . These specializations are irreducible over Kτ as well, since Kτ is a regular extension of Kα by [FJ08, Corollary 2.6.5(d)]. Remark 5.5. Padmavathi Srinivasan has constructed a field K with stronger properties than the L∞ in the proof of Theorem 1.5, namely that C(K) is finite for every K-curve C of absolute genus > 1, while there is a degree 2 field extension L ⊃ K that is ample [Sri15]. Acknowledgements We thank the referee for many insightful comments and corrections. References [BSF13] Lior Bary-Soroker and Arno Fehm, Open problems in the theory of ample fields, Geometric and differential Galois theories, Sémin. Congr., vol. 27, Soc. Math. France, Paris, 2013, pp. 1–11 (English, with English and French summaries). MR3203546 4 [BLR90] Siegfried Bosch, Werner Lütkebohmert, and Michel Raynaud, Néron models, Ergebnisse der Mathematik und ihrer Grenzgebiete (3) [Results in Mathematics and Related Areas (3)], vol. 21, SpringerVerlag, Berlin, 1990. MR1045822 (91i:14034) [FJ08] Michael D. Fried and Moshe Jarden, Field arithmetic, 3rd ed., Ergebnisse der Mathematik und ihrer Grenzgebiete. 3. Folge. A Series of Modern Surveys in Mathematics [Results in Mathematics and Related Areas. 3rd Series. A Series of Modern Surveys in Mathematics], vol. 11, Springer-Verlag, Berlin, 2008. Revised by Jarden. MR2445111 (2009j:12007) [Jar11] Moshe Jarden, Algebraic patching, Springer Monographs in Mathematics, Springer, Heidelberg, 2011. MR2768285 (2012e:12007) [JK10] Markus Junker and Jochen Koenigsmann, Schlanke Körper (slim fields), J. Symbolic Logic 75 (2010), no. 2, 481–500, DOI 10.2178/jsl/1268917491. MR2648152 (2011f:03047) [Koe01] Jochen Koenigsmann, Solvable absolute Galois groups are metabelian, Invent. Math. 144 (2001), no. 1, 1–22, DOI 10.1007/s002220000117. MR1821143 (2002a:12006) [Pop96] Florian Pop, Embedding problems over large fields, Ann. of Math. (2) 144 (1996), no. 1, 1–34. MR1405941 (97h:12013) [Sam66] Pierre Samuel, Compléments à un article de Hans Grauert sur la conjecture de Mordell, Inst. Hautes Études Sci. Publ. Math. 29 (1966), 55–62 (French). MR0204430 (34 #4272) [Sri15] Padmavathi Srinivasan, A virtually ample field that is not ample, 2015. Preprint. School of Mathematics, Tel Aviv University, Ramat Aviv, Tel Aviv 6139001, Israel E-mail address: jarden@post.tau.ac.il Department of Mathematics, Massachusetts Institute of Technology, Cambridge, MA 02139-4307, USA E-mail address: poonen@math.mit.edu URL: http://math.mit.edu/~poonen/ 5